Statistical Dominance Algorithm

SDA performs the initial stage of edge detection or segmentation of digital grayscale images. The algorithm counts the number of pixels with a given relation to the central point of the neighborhood. The output image is a statistical result of the dominance of points over their neighborhoods and allows the classification of these points to be determined (peak, valley, and slope). Therefore, this solution allows the impact of noise or uneven illumination in image results to be reduced. [1], PDF

Main applications of the algorithm:

Download:

The main idea

for (x = N; x < SX - N; x++)
  for (y = N; y < SY - N; y++)
  {
    imgout[x,y] = 0;	 
 
    for (i = -N; i <= N; i++)
      for (j = -N; j <= N; j++)
        if (i * i + j * j <= R * R)
          if (imgin[x + i, y + j] >= imgin[x, y] + threshold)
            imgout[x, y]++;
  }

source	SDA (r=50, threshold=50)	SDA (r=50, threshold=50), binarization threshold=150

	SDA (r=25, threshold=14)	plugin view

Segmentation

Segmentation of heart chambers (atria, ventricles) and valve leaflets on CT data

8-bit	SDA	settings	16 bit (DICOM) -> PNG 16bit 1:1		SDA	settings

---

Edge detection

Wood	grayscale	Red	Blue	SDA of grayscale	SDA of Red	SDA of Blue	SDA of grayscale
neighborhood:				horizontal	horizontal	horizontal	disc

SDA: r=20, threshold=0, neighborhood=horizontal

Normalization: Magnetic Resonance Imaging [3]

Normalization: Corneal Endothelium Images [1]

source	normalized	smoothing	SDA	SDA
		maximal filter 3x3	r=2, threshold=0	r=3, threshold=0

Enchancing the trabecular structure of bone on the X-ray images [7]

source	histogram equalisement	CLAHE	CLAHE+hist. equal.	SDA (r=50, thr=0, neg, disc)
		maximal filter 3x3	r=2, threshold=0	r=3, threshold=0

Car Licence Plates

Car	Car - SDA, r=5.0, threshold=30	Car - SDA, r=5.0, threshold=120

---

References: